Cloud Penetration Testing
Fail
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: HIGHCOMMAND_EXECUTIONDATA_EXFILTRATIONPERSISTENCECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches official CLI tools and configuration scripts from well-known service providers including Amazon Web Services and Google Cloud Platform. It also installs standard security auditing tools such as ScoutSuite and Pacu via Python package management.
- [COMMAND_EXECUTION]: Utilizes
Invoke-AzVMRunCommandandgcloud beta compute sshto execute commands and scripts on remote virtual machine instances within cloud environments. - [DATA_EXFILTRATION]: Provides instructions for extracting sensitive data, including secrets from Azure Key Vaults, environment variables from AWS Lambda functions, and credential tokens from cloud provider metadata services (169.254.169.254).
- [PERSISTENCE]: Documents methods for establishing persistent access, such as creating backdoor service principals with 'Owner' roles, generating new AWS access keys for existing users, and creating administrative accounts via command-line interfaces.
- [CREDENTIALS_UNSAFE]: Suggests practices that could lead to credential exposure, such as exporting authentication contexts to temporary file system locations (e.g.,
C:\Temp\AzureAccessToken.json) and retrieving plaintext secrets from memory using PowerShell interop. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to its interaction with untrusted cloud environment data.
- Ingestion points: Reads data from external sources including cloud resource inventories, API metadata responses, and instance configurations (SKILL.md, advanced-cloud-scripts.md).
- Boundary markers: None identified; instructions do not specify the use of delimiters or 'ignore' instructions for external data.
- Capability inventory: Possesses extensive capabilities including remote command execution, file system modifications, and network operations across three major cloud platforms.
- Sanitization: There are no documented procedures for sanitizing or validating data retrieved from cloud environments before it is processed or acted upon.
Recommendations
- HIGH: Downloads and executes remote code from: https://sdk.cloud.google.com - DO NOT USE without thorough review
Audit Metadata